Agent-based AI is becoming a key element in the value chain
5 MAY, 2026
Marius Wennersten, Portfolio Manager at DNB Asset Management
After several years of intensive investment in AI, the technology sector is entering a new phase. The cycle is expanding, and the risk-return profile along the value chain is shifting as the focus moves from capacity expansion to actual utilisation and monetisation at the higher levels of the technology platform.
The emergence of agent-based AI is a key driver of this transformation. We are now seeing clear signs of accelerated adoption, not just investment. Recurring revenue at leading providers such as OpenAI and Anthropic has risen significantly in a short space of time, indicating a surge in activity as AI agents are increasingly deployed to perform real-world tasks.
The longer-term opportunities remain substantial. Global expenditure on the workforce far exceeds expenditure on data centres and software, illustrating the scale of the efficiency potential that AI could unlock over time. However, the path to monetisation is unlikely to be linear, and value creation will not be evenly distributed.
Agent-based AI shifts focus to companies further up the value chain
To date, returns have been concentrated in infrastructure. Semiconductor and hardware providers have been the main beneficiaries of the investment cycle, with AI infrastructure stocks significantly outperforming their peers. Consequently, in our view, valuations for infrastructure reflect significantly more ‘AI optimism’ than those for the platform and application layers above.
We expect agentic AI to be a key mechanism through which value shifts further up the hierarchy. AI agents need to be coordinated, deployed and integrated into business processes, and many companies are choosing to do this via their existing cloud and application partners, where data, computing power, security and identity management are already embedded. Provided this pattern continues, the value generated by the agentic shift is likely to benefit not only the infrastructure layer but also the platform and application layers.
We believe that the three largest cloud platforms – Amazon Web Services, Microsoft Azure and Google Cloud – are well positioned. Cloud growth has accelerated again in recent quarters as earlier investments are now being converted into revenue. These platforms benefit from close corporate relationships and operational leverage, whilst retaining the flexibility to reallocate capacity should AI-related spending slow down. We continue to view the risk-reward profile in this segment as attractive.
Software: under pressure, but not finished
Software remains a much-discussed part of the sector. The SaaS segment has underperformed the broader indices significantly over the past year, due to a combination of structural concerns and the lack of a positive turnaround in fundamentals.
Some of the structural concerns are justified. Disintermediation, bundling and lower switching costs can put pressure on established providers, particularly as new architectures reshape the way problems are solved. Customer service is one example: how this function is delivered in 2030 is likely to differ significantly from 2020, and AI-native market participants could be well-positioned to capitalise on parts of this shift. Nevertheless, established companies retain advantages through embedded data, integrations and process depth.
We also believe that parts of the argument are exaggerated. The notion that AI-assisted programming undermines the value of enterprise software overlooks where that value actually lies. In most cases, it does not lie in code generation, but in domain expertise, workflow integration and the codification of complex business processes. Companies such as SAP derive their economic value from process depth and their embedding in customer workflows, not from their ability to write code.
Against this backdrop, the sell-off is creating areas where valuations appear excessively low relative to fundamentals. In some cases, high-quality software companies are trading at low single-digit revenue multiples, despite strong gross margins, healthy growth and the potential for a significant expansion of the operating margin over time.
Transition from the build phase to the adoption phase
The AI cycle is transitioning from a build phase to an adoption phase. The focus is shifting from infrastructure expansion to monetisation across the entire stack, and we see the most attractive opportunities where this shift is still underestimated.
We remain constructive on the technology sector, supported by its historical ability to deliver above-average earnings growth and drive productivity gains across the wider economy. We view agentic AI as the next stage of this productivity cycle. At the same time, the valuation range within the sector is wide, which is why our approach is constructive but disciplined. We focus on companies where long-term earnings power and cash generation appear undervalued, and avoid areas where expectations and valuations appear excessively high.
